Identification and Evaluation of Freight Demand Factors (2012)

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

4. Conclusions and Recommendations Forecasting the demand for freight transportation in our nation is crucial for many segments of our economy. Over the long term, forecasts support planned investments in infrastructure. Over the short term, forecasts identify challenges and opportunities to operators of transportation services and investors in transportation equipment. A nd because the various enterprises that provide freight transportation services require continued capital investment, operational enhancements, and improvements in technology and communication to ensure their existence, gaining an understanding of possible future demand is highly prized. This research began by examining previous studies and other literature on factors influencing freight transportation demand. The research team sought to understand what factors affect the “pure” demand for transportation – the amount of tons of goods extracted, grown, produced and then utilized or consumed – vs. “network” factors that influence the timing, location, and distance goods are moved. A wide variety of potential independent influencing variables were considered that affect either these “pure” or direct freight demand factors, or the “network” or indirect factors. The analysis presumed that freight demand was independent by mode and, with only one notable exception, this proved to be the right choice. The research team placed several constraints on the data that would be used to measure the influences on freight transportation demand. The data had to be regularly and freely available during the 28-year study period, which limited the ultimate selection. Expecting that a predictive model would apply for the entire time period might have been a strong assumption as suggested by sub-period correlations. However, the example backcasts may suggest otherwise. There have been significant shifts in the way the same economic and demographic data affects demand for freight transportation, indicating that the predictive models developed might be best suited for short-to-medium-term time periods. Although a wide variety of prospective independent influences were originally considered, the variables ultimately chosen were primarily economic measurements tallied by the United States government. Few of the considered variables, especially the trucking data, had reliable information over time periods shorter than a year, limiting the team’s ability to test a variety of time-lagged influences, for example by month or quarter-year. In subsequent research, it might be helpful to remove the constraint that utilized data must be widely and freely available. Information on w arehouse development, housing patterns, transportation equipment costs, etc. might be difficult to obtain and isolate, but it might also help clarify some of the relationships that generate freight transportation growth. The ability to “time lag” the independent factors provided some of the most interesting and beneficial results. Both the Purchasing Manager’s Index and Housing Starts apparently had significant effects on subsequent freight transportation demands from year-to-year. Moreover, the models confirmed that the years following NAFTA saw a substantial increase in new freight transportation demand. 56

The regression analysis models indicated very good fit, with R2values above .90 and a relatively parsimonious set of independent factors, across the various measures of freight demand. Rail car loadings have typically been a good historical indicator of industrial activity, and vice-versa as it turned out. Freight transportation plays a central role in the movement of materials and goods in our industrial as well as consumption economy, so this “fit” was not surprising. There were observable differences that matched our intuitive understanding of what affects freight demand. M easures of “pure” industrial production and personal consumption affected rail tons while trade, imports and exports, because of their lengths-of-haul to and from seaports, affected both truck and rail ton-miles. M easures of inland waterway freight were more problematic to model since use of this mode has actually dropped during the past 20 years, even as the economy has mostly grown. Waterway ton-miles was the one dependent variable that showed very high negative correlation with a substitutable alternative – rail ton-miles. Early warning indicators were identified by noting if and how well fluctuations in independent factors’ values in one time period predicted changes in freight demand during the following time period. Depending on the freight transportation demand being measured, Purchasing Manager’s Index, Number of Households (i.e., “Housing Starts”) and even Fuel Prices, all showed some ability to predict freight transportation demand in subsequent periods, typically in conjunction with other economic variables. The predictive value of these “Early Warning” variables is even more helpful if they explain a large portion of changes in future freight demand. Selected summary results from our Regression analysis include the following excerpts from the predictive models that show what the ∆Change in Freight vs. 10% Change in Indepe ndent Variable would be for the two most primary independent influencing variables in one of the top predictive models of freight demand: Freight Demand Primary Influencing Variable ∆Freight / 10% Change Secondary Influencing Variable ∆Freight/ 10%Change Rail Tonnage Industrial Prod. Index 8.4% Trade Wghtd. Index (Broad Currencies) - 1.4% Rail Ton-Miles Industrial Prod. Index 9.6% Inventory/Sales Ratio - 4.7% Rail Train-Miles GDP in Real$ 5.7% Purchasing Mgr’s Index (Lagged from prior yr) 1.8% Rail Car-Miles GDP in Real$ 6.6% NAFTA – two yrs following 0.4% Rail Rev Ton-Miles GDP in Real$ 10.6% NAFTA – two yrs following 0.6% Truck Ton-Miles Total Trade in Real$ 1.0% Gasoline Price - 0.5% Truck Vehicle Miles Total Trade in Real$ 1.0% Inventory/Sales Ratio - 1.7% Water Tonnage Total Capacity Utiliz. 8.6% Grain+Coal Tonnage 0.9% Water Ton-Miles Rail Ton-Miles - 4.4% IWTF Gas Tax (Lagged from prior yr) - 1.3% Out of concern that many of the independent influencing variables were closely correlated with each other, the team sought an additional method to overcome potential bias in the basic 57

regression analysis. Principal Component Analysis (PCA) provided a helpful method to combine the explanatory benefit of multiple, similar independent economic factors on resulting freight transportation demand. While they are less intuitive and difficult to deconstruct, the method has been proven to be useful elsewhere and may help develop accurate predictive models even when mostly highly collinear independent data is all that is available. PCA models do not provide good correspondence between changes in the independent variables with changes in the dependent measures of freight transportation since the construction of the components affects this correspondence. Still, PCA models developed by the research team for this project were able to explain most of the variability in the truck and rail demand measures, and performed well in a backcasting exercise, accurately matching predicted with observed data. Recommendations for Additional Analysis • Extend the data capture through the 2008-2011 time period to determine if the same relationships exist during our current recession. • Experiment with more or different independent influencing variables and freight demand measurements that might be available on a m onthly or quarterly basis, even if fees were involved or the data were available for shorter (5- to15-year) time periods. • Consider regional measurements of economic and other freight-influencing data such as state GDP, housing starts and imports/exports by regional port. C ompare with state-based measures of truck and rail traffic to determine whether similar predictive relationships exist on a regional basis. • Evaluate new and emerging independent variables such as the growth in modern warehouse real estate development and technologies or the shift in U.S. agricultural exports towards the Pacific Rim 58